Move RB info from receiver reports into the internal source that the RR
are about, and deprecate (but retain) the old mapping where each
external source has only a single RB entry in the rtp statistics.
The old method is broken if a remote peer uses a single ssrc to send
receiver reports for more than one of our internal sources, other
as multiple RB in a single packet, or alternate RB in different reports.
In each case only the most recent entry was kept, overwriting data for
other internal sources.
In multicast scenarios each internal source may receive multiple
receiver reports from different peers. To support that, all received
RR's are now stored into a hash table indexed by the sender's SSRC,
and all RRs are placed into an array when generating statistics, so
that the information from all peers is retrievable.
The current deficient behaviour (adding RB info into non-internal RTPSources) is
deprecated but kept in order to be backward compatible, and retained
that way in the generated statistics structure.
Refs
[1] https://tools.ietf.org/html/rfc3550#section-6.4.1
Based on a patch by Fede Claramonte <fclaramonte@twilio.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7424>
A zero-sized box is not really a problem and can be skipped to look at any
possibly following ones.
BMD ATEM devices specifically write a zero-sized bmdc box in the sample
description, followed by the avcC box in case of h264. Previously the avcC box
would simply not be read at all and the file would be unplayable.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7564>
Timestamps are untouched by default, but the new mode can now be enabled to replace RTP timestamps
with ones generated from the buffer PTS. Making it an enum in case different modes are needed in the future.
That allows for a rtpjitterbuffer to do proper drift compensation, so that the stream coming out of gst-rtsp-server
is not drifting compared to the pipeline clock and also not compared to the RTCP NTP times.
Most of the code is borrowed from rtpbasepayload, as it's exactly its behaviour which I wanted to bring here.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7526>
It has to be included in the block duration but in GStreamer we're not
including it in the buffer duration, so it has to be added again here.
Not including it in the block duration can lead to fatal errors when playing
back with Firefox if there are more padding samples than actual samples, e.g.
> D/MediaDemuxer WebMDemuxer[7f6a0808b900] ::GetNextPacket: Padding frames larger
> than packet size, flagging the packet for error (padding: {13500000,1000000000},
> duration: {6000,1000000}, already processed: false)
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7502>
By setting the earliest time to timestamp + 2 * diff there would be a difference
of 1 * diff between the current clock time and the earliest time the element
would let through in the future. If e.g. a frame is arriving 30s late at the
sink, then not just all frames up to that point would be dropped but also 30s of
frames after the current clock time.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7459>
splitmuxsink can't possibly know how much latency it will introduce as it always
keeps one GOP around before outputting something. This breaks the latency
configuration of the pipeline and we're better off just pretending that
everything downstream of the sinkpads is not live.
Especially muxers that are based on aggregator and time out on the latency
deadline can easily misbehave otherwise as the deadline will be exceeded usually.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7499>
If two (or more) rtpfunnel elements are cascaded, then only one will
realistically have information on the particular ssrc that is in use for a
particular input stream. As such, any key unit requests may never reach the
corresponding encoder.
This has been discovered by combining simulcast and BUNDLE with webrtcbin.
simulcast uses one rtpfunnel, and BUNDLE uses another rtpfunnel.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7405>
Parts may emit bus messages that want to take the splitmuxsrc
lock and prevent the downward state change. Avoid a deadlock
after a part sends an error message by taking a ref and
dropping the lock around the unprepare call
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7053>
Publish fragment-id in the messages that splitmuxsink and splitmuxsrc
send, so when they are received out of order (due to async finalization,
for example), they can still be identified / ordered correctly.
Fix a race in the splitmuxsink unit test where messages might be
received out of order
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7053>
Add a `num-lookahead` property that will 'prepare' a number of
fragments in advance of the playhead if they have been deactivated
or closed by a limited number of `num-open-fragments`. It can help
to avoid any play stalls reading the indexes or headers of the next
file from high-latency media or on resource limited machines.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7053>
Publish the playback offset for and duration into the
splitmuxsink-fragment-closed bus message as each fragment
finishes.
These can be passed to splitmuxsrc via the 'add-fragment'
signal to avoid splitmuxsrc measuring all files on startup
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7053>
Add a reasonably large default for the number of simulataneous
files to open, that won't affect users that split recordings into
a few large files, but will help prevent fd exhaustion for users
that make recordings with lots of small fragments
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7053>
When calculating the timestamp offset to apply to
media streams in a fragment, ensure that all fragments
are offset "together" to preserve alignment in cases
where there might gaps in a recording at a fragment boundary.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7053>
Add a signal that allows adding fragments with a specific offset
and duration directly to splitmuxsrc's list. By providing the
fragment's offset on the playback timeline and duration directly,
splitmuxsrc doesn't need to measure the fragment making for faster
startup times.
Add a bus message that's published when fragments are measured,
reporting the offset and duration, so they can be cached by an
application and used on future invocations.
Add examples for handling the bus message and using the 'add-fragment'
signal.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7053>
Add a property to limit the number of parts splitmux will open
simultaneously. Modify the part handling to support deactivating
and reactivating the demuxing for each part.
The default is '0', to preserve the existing behaviour of opening
all parts at the beginning.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7053>
Not doing so would mean that tags would be overidden by any tag events sent by
upstream. Also only send a tag event directly if upstream never sent one.
By default use GST_TAG_MERGE_REPLACE to override tags that exist in both the
upstream event and this element with the ones from this element, but provide a
new "merge-mode" property to adjust the behaviour.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7145>
Specify "layout" field in src template to make sure it's
set and gets fixated properly if the downstream element
supports both interleaved and non-interleaved caps.
Fixes
gst_pad_set_caps: assertion 'caps != NULL && gst_caps_is_fixed (caps)' failed
critical with e.g.
gst-launch-1.0 rtpdtmfsrc ! rtpdtmfdepay ! audioconvert ! fakesink
Not that the layout really matters in our case since we always
output mono anyway, but non-interleaved requires adding AudioMeta,
so this is the easiest fix.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7036>
Upon fatal errors the loop function will first post an error message
then push out an EOS event.
An application may react immediately to the error message by setting the
state of the pipeline to NULL, meaning by the time we push out the EOS
event PAUSED_TO_READY may have reset the seek seqnum to -1.
While this is harmless, the assertion when setting an invalid seqnum
isn't tidy, fix this by simply not resetting to INVALID as it serves no
practical purpose and the next READY_TO_PAUSED will select a new seqnum
anyway.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7032>
Even if no new synchronization information is available.
This is necessary because the timestamp offset logic in rtpbin depends
on the base RTP time that is determined by the jitterbuffer, but this
changes all the time (especially in mode=slave) and the timestamp
offsets have to be updated accordingly. Doing so is especially important
if they're only determined by the RTP-Info, which never changes from the
very beginning.
The interval can be configured via the new min-sync-interval property.
Synchronization happens at least that often, but at most as often as the
old sync-interval property allows.
Both intervals are now based on the monotonic system clock.
Additionally, clean up synchronization code a bit, only emit either
inband NTP or RTCP SR synchronization at the same time, based on which
one has the more recent time information, and only emit RTP-Info
synchronization if it wasn't provided previously at the same time as the
NTP-based synchronization information.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
There is generally no requirement to ignore RTCP SR if the RTP time of
the SR differs a lot from the last received RTP packet. The mapping
between RTP and NTP time stays valid until there was a stream reset, in
which case we wouldn't use that information anyway.
When using rtcp-sync-send-time=false the default of 1s difference can
easily be exceeded, e.g. if encoding of the stream after capture adds
more than 1s of latency.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
Never is useful for some RTSP servers that report plain garbage both via
RTCP SR and RTP-Info, for example.
NTP is useful if synchronization should only ever happen based on RTCP
SR or NTP-64 RTP header extension.
Also slightly change the behaviour of always/initial to take RTP-Info
based synchronization into account too. It's supposed to give the same
values as the RTCP SR and is available earlier, so will generally cause
fewer synchronization glitches if it's made use of.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
Instead of switching on the very first stream, require that all streams
have switched before switching to the different synchronization
mechanism.
Without this there will be a noticeable gap during the switch. E.g. when
going from RTP-Info to NTP-based association, first the first stream
only would get an offset, then the first two, ... then all of them.
Depending on the order of streams this will cause a lot of changes in
ts-offset during the transition.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
Previously these parameters were randomly changed in the body of the
function to avoid having to declare a new variable, which made the code
very hard to follow. By marking them as const this won't be possible
anymore in the future.
Also the RTP clock-base (RTP time from RTSP RTP-Info) is an unsigned
64 bit integer as it's an extended RTP timestamp.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
Both were entangled previously and very hard to follow what happens
under which conditions. Now as a very first step the code decides which
of the two cases it is going to apply, and then proceeds accordingly.
This also avoids calculating completely invalid values along the way and
even printing them int the debug output.
Also improve debug output in various places.
This shouldn't cause any behaviour changes.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
This simplifies the code as it's a much simpler case than the normal
inter-stream synchronization, and interleaving it with that only
reduces readability of the code.
Also improve some debug output in this code path.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
If we have constant duration buffers, set the duration on
outgoing buffers, like rtpmp4adepay does. This fixes
problems with (for example) muxers like mp4mux not writing
the duration of the final sample into the index.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6878>
During RTP-Info synchronization, clock_base was temporarily switched
from the actual clock-base to the base RTP time and then back some lines
later.
Instead directly work with the base RTP time. The comment about using a
signed variable for convenience doesn't make any sense because all
calculations done with the value are unsigned.
Similarly, rtp_clock_base was overridden with the rtp_delta when
calculating it, which was fine because it is not used anymore
afterwards. Instead, introduce a new variable `rtp_delta` to make this
calculation clearer.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6536>
It's not in the same period as the current RTP base time but always in
the very first period. This avoids using it again at a much later time.
The code in question is only triggered with rtcp-sync=rtp-info.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6536>
It is compared to other extended RTP timestamps all over rtpjitterbuffer
and since 4df3da3bab the initial extended RTP timestamp is not equal
anymore to the plain RTP time.
Continue passing a non-extended RTP timestamp via the `sync` signal for
backwards compatibility. It will always be a timestamp inside the first
extended timestamp period anyway.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6536>
When the buffer DTS is estimated based on arrival time at the
jitterbuffer (rather than provided on the incoming buffer itself),
it shouldn't be used for skew adjustment. The typical case is
packets being deinterleaved from a tunnelled TCP/HTTP RTSP stream,
and the arrival times at the jitter buffer are not well enough
correlated to usefully do skew adjustments.
This restores the original intended behaviour for the 'estimated dts'
path, that was broken years ago during other jitterbuffer refactoring.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6509>
If we can calculate timestamps for buffers, then set the duration
on outgoing buffers based on the number of samples depayloaded.
This can fix the muxing to mp4, where otherwise the last packet
in a muxed file will have 0 duration in the mp4 file.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6447>
When we're doing a state change from PLAYING to NULL, first we invoke
gst_rtspsrc_loop_send_cmd_and_wait (..., CMD_CLOSE, ...) during
PAUSED_TO_READY which will schedule a TEARDOWN to happen async on the
task thread.
The task thread will call gst_rtspsrc_close(), which will send the
TEARDOWN and once it's complete, it will call gst_rtspsrc_cleanup()
without taking any locks, which frees src->streams.
At the same time however, the state change in the app thread will
progress further and in READY_TO_NULL it will call gst_rtspsrc_stop()
which calls gst_rtspsrc_close() a second time, which accesses
src->streams (without a lock again), which leads to simultaneous
access of src->streams, and a segfault.
So the state change and the cleanup are racing, but they almost always
complete sequentially. Either the cleanup sets src->streams to NULL or
_stop() completes first. Very rarely, _stop() can start while
src->streams is being freed in a for loop. That causes the segfault.
This is unlocked access is unfixable with more locking, it just leads
to deadlocks. This pattern has been observed in rtspsrc a lot: state
changes and cleanup in the element are unfixably racy, and that
foundational issue is being addressed separately via a rewrite.
The bandage fix here is to prevent gst_rtspsrc_stop() from accessing
src->streams after it has already been freed by setting src->state to
INVALID.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6302>
The transport stream only returned the CAPS for the first matching PT entry
from the `ptmap`. Other SSRC with the same PT where not included. For a stream
which bundled multiple audio streams for instance, only the first SSRC was
knowed to the SSRC demux and downstream elements.
This commit adds all the `ssrc-` attributes from the matching PT entries.
The RTP jitter buffer can now find the CNAME corresponding its SSRC even if it
was not the first to be registered for a particular PT.
The RTP PT demux removes `ssrc-*` attributes cooresponding to other SSRCs
before pushing SSRC specific CAPS to downstream elements.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6119>
And also re-timestamp them with the current buffer's PTS.
Not doing so keeps the timestamps of event packets as
GST_CLOCK_TIME_NONE or the timestamp of the previous buffer, both of
which are bogus.
Making sure that (especially) the first packet has a valid timestamp
allows putting e.g. the NTP timestamp RTP header extension on it.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5173>
Parse the speed and scale in the server's response
*before* the range, so that the range start/stop
are swapped (or not swapped) correctly based
on the server's actual chosen values. Otherwise,
the old rate from the segment is used - what the
last seek asked for, but not necessarily what
the server chooses.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6248>
After a flushing seek, rtspsrc doesn't reset the last_ret value for
streams, so might immediately shut down again when it resumes pushing
buffers to pads due to a cached `GST_FLOW_FLUSHING` result
Prevent a stored flushing value from immediately stopping
playback again by resetting pad flows before (re)starting
playback.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6137>
Because this depayloader may build several output buffers within one
process run we push them all into a GstBufferList and push them out at
once to make sure that each buffer gets notified about each header
extension.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5378>
Because this depayloader may build several output buffers within one
process run we push them all into a GstBufferList and push them out at
once to make sure that each buffer gets notified about each header
extension.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5378>
